Python Machine Learning Using Scikit Learn, TensorFlow, PyTorch, and Keras, an Introductory Journey into Machine Learning,… (Vere salazar) (Z-Library)

  Python machine learning   Using Scikit Learn, TensorFlow, PyTorch, and Keras, an Introductory Journey into Machine Learning, Deep Learning, Data Analysis, Algorithms, and Data Science Vere salazar      

© Copyright 2024 by vera poe all rights reserved.   The content contained within this book may not be reproduced, duplicated or transmitted without direct written permission from the author or the publisher.   Under no circumstances will any blame or legal responsibility be held against the publisher, or author, for any damages, reparation, or monetary loss due to the information contained within this book, either directly or indirectly.   Legal notice:   This book is copyright protected. It is only for personal use. You cannot amend, distribute, sell, use, quote or paraphrase any part, or the content within this book, without the consent of the author or publisher.   Disclaimer notice:   Please note the information contained within this document is for educational and entertainment purposes only. All effort has been executed to present accurate, up to date, reliable, complete information. No warranties of any kind are declared or implied. Readers acknowledge that the author is not engaging in the rendering of legal, financial, medical or professional advice. The content within this book has been derived from various sources. Please consult a licensed professional before attempting any techniques outlined in this book.   By reading this document, the reader agrees that under no circumstances is the author responsible for any losses, direct or indirect, that are incurred as a result of the use of information contained within this document, including, but not limited to, errors, omissions, or inaccuracies.  

Table of contents Chapter 1: machine learning: a brief history Donald hebb - the organization of behavior Samuel arthur - neural networks, checkers and rote learning Rosenblatt’s perceptron Marcello pelillo - the nearest neighbor algorithm Perceptrons and multilayers Going separate ways Robert schapire - the strength of weak learnability Advancing into speech and facial recognition Present day machine learning Chapter 2: fundamentals of python for machine learning What is python? Why python? Other programming languages Effective implementation of machine learning algorithms Mastering machine learning with python Chapter 3: data analysis in python Importance of learning data analysis in python Building predictive models in python Python data structures Python libraries for data analysis Chapter 4: comparing deep learning and machine learning Deep learning vs machine learning Problem solving approaches

Different use cases Chapter 5: machine learning with scikit-learn Representing data in scikit-learn Features matrix Target arrays Estimator api Supervised learning in scikit-learn Unsupervised learning in scikit-learn Chapter 6: deep learning with tensorflow Brief history of tensorflow The tensorflow platform Tensorflow environments Tensorflow components Algorithm support Creating tensorflow pipelines Chapter 7: deep learning with pytorch and keras Pytorch model structures Initializing pytorch model parameters Principles supporting keras Getting started Keras preferences Keras functional api Chapter 8: role of machine learning in the internet of things (iot) Chapter 9: looking to the future with machine learning The business angle

Ai in the future Conclusion

Introduction The, me,ntion of de,ve,lope,rs and programming usually has a lot of pe,ople, dire,cting the,ir thoughts to the, wide,r study of compute,r scie,nce,. compute,r scie,nce, is a wide, are,a of study. in machine, le,arning, compute,rs le,arn from e,xpe,rie,nce,, aide,d by algorithms. to aid the,ir cause,, the,y must use, data with spe,cific fe,ature,s and attribute,s. this is how the,y ide,ntify patte,rns that we, can use, to he,lp in making important de,cisions. in machine, le,arning, assignme,nts are, groupe,d unde,r diffe,re,nt cate,gorie,s, such as pre,dictive, mode,ling and cluste,ring mode,ls. the, conce,pt be,hind machine, le,arning is to provide, solutions to pe,rtine,nt proble,ms without ne,ce,ssarily waiting for dire,ct human inte,raction. Machine, le,arning and artificial inte,llige,nce, today are, the, re,ality that we, dre,amt of ye,ars ago. the,se, conce,pts are, no longe,r confine,d to fictional ide,as in movie,s, but the,y have, be,come, the, backbone, of our daily live,s. if you think about your inte,rne,t activity all through the, day, you inte,ract with machine, le,arning mode,ls all the, time,. how many time,s have, you had a we,bsite, translate,d from a fore,ign language, to your native, language,? think about the, numbe,r of time,s you have, be,e,n assiste,d through a chatbot, or use,d facial and voice, re,cognition programs. all the,se, are, instance,s whe,re, we, inte,ract with machine, le,arning mode,ls, and the,y he,lp by making our live,s e,asie,r. Like, any othe,r discipline,, machine, le,arning doe,s not e,xist in isolation. many conce,pts in machine, le,arning are, inte,rtwine,d with de,e,p le,arning and artificial inte,llige,nce,. the,re, are, othe,r subje,cts that share, similaritie,s with machine, le,arning, but for the, purpose, of this book, we, will focus on de,e,p le,arning and artificial inte,llige,nce,. This be,ing the, first book in a se,rie,s of e,nlighte,ning books about machine, le,arning, will introduce, you to the, fundame,ntal ide,ologie,s you should

unde,rstand the, te,chnology, syste,ms, and proce,dure,s use,d in machine, le,arning, and how the,y are, conne,cte,d. Artificial inte,llige,nce, branche,s off from machine, le,arning, but the,y share, a lot of similaritie,s. tracing the,se, two studie,s back in time,, the,y share, the, same, path for most of the,ir history. while, machine, le,arning focuse,s on building mode,ls that le,arn through algorithms and can ope,rate, without human inte,rve,ntion, artificial inte,llige,nce, focuse,s on simulating human e,xpe,rie,nce,s and inte,llige,nce, through computing. it is safe, to say that machine, le,arning is a subclass of artificial inte,llige,nce, be,cause, we, work towards building machine,s that can simulate, human de,cision-making proce,sse,s, albe,it by le,arning through data. De,e,p le,arning introduce,s us to anothe,r division of machine, le,arning whe,re, artificial ne,ural ne,tworks (ann) are, e,mploye,d in making important de,cisions. in de,e,p le,arning, the, ne,ural ne,tworks use, laye,re,d structure,s whose, functions are, similar to the, functions of a he,althy human brain. the,re,fore,, machine, le,arning, de,e,p le,arning, and artificial inte,llige,nce, are, thre,e, discipline,s that are, inte,rconne,cte,d in more, ways than one,. whe,n you commit to le,arning one, of the,m, you will inadve,rte,ntly have, to le,arn about the, othe,rs too at some, point. In machine, le,arning, de,e,p le,arning is a cate,gory that focuse,s on using algorithms to e,mpowe,r syste,ms and build mode,ls that are, similar in ope,ration to the, human brain. the, pre,se,nt e,xcite,me,nt and hype, around de,e,p le,arning come,s from the, fundame,ntal studie,s in ne,ural ne,tworks. re,se,arch in ne,ural ne,tworks has be,e,n carrie,d out for many ye,ars, and could date, back longe,r than the, history of machine, le,arning. this is be,cause, part of this knowle,dge, is e,mbe,dde,d in ne,urological studie,s without an iota of re,fe,re,nce, to machine, le,arning or computing. The,re, have, be,e,n major stride,s in machine, le,arning re,se,arch ove,r the, ye,ars, e,spe,cially with re,spe,ct to de,e,p le,arning. while, we, must re,cognize, the, scalability of the,se, discipline,s, the, advance,me,nt in the,se, te,chnologie,s

is made, possible, by thre,e, important factors; the, de,ve,lopme,nt of e,fficie,nt algorithms, the, incre,asing and matching de,mand for significant computing re,source,s, and the, incre,ase, in the, inte,rne,t population, he,nce, massive, chunks of data are, available, to train and e,mpowe,r the,se, machine,s. So how do we, find the, link be,twe,e,n de,e,p le,arning and machine, le,arning? the, answe,r lie,s in how the,se, mode,ls ope,rate,. from a basic pe,rspe,ctive,, you work with mode,ls which re,ce,ive, pre,de,fine,d input and output data. input data could be, anything from te,xt instructions, to nume,rical input, or audio, vide,o, and image,s in diffe,re,nt me,dia formats. base,d on the, input, the, spe,cific mode,l you use, will the,n de,rive, an output that me,e,ts your instructions. output could be, anything from ide,ntifying an individual’s name, to de,fining the,ir tribe,. the, corre,ct answe,r de,pe,nds on the, kind of input data you provide, the, machine, le,arning mode,l. As you le,arn about the,se, ne,tworks, you must also spare, time, to sharpe,n your data analysis and data handling skills. one, skill you must be, good at is how to pre,pare, data, e,spe,cially how to cle,an data. machine, and de,e,p le,arning mode,ls de,pe,nd on data for accuracy. inaccuracie,s in the, input data will affe,ct the, output. many mistake,s happe,n at data e,ntry and if the,se, are, not che,cke,d, you will e,nd up with a good machine, le,arning mode,l that cannot de,live,r the, outcome, e,xpe,cte,d. this is why data cle,aning, and data analysis in ge,ne,ral, are, important proce,sse,s. Once, your mode,l has sufficie,nt data, it should pre,dict outcome,s according to the, input provide,d and the, instructions upon which the, mode,l trains. today the,re, are, many machine, le,arning mode,ls that are, alre,ady in use,, including te,xtcnn, yolo, ince,ption, and face,ne,t. An ove,rvie,w of machine, le,arning make,s it sound like, a simple, conce,pt. for those, who have, programme,d for ye,ars in this fie,ld, it ge,ts e,asie,r ove,r time,. while, the, machine, le,arns, you also ne,e,d to ke,e,p le,arning, so you are, in a be,tte,r position to furthe,r your skill in machine, le,arning. at a be,ginne,r le,ve,l, knowle,dge, of the, basic conce,pts should se,t you on the, right path.

Anothe,r important conce,pt you should ne,ve,r forge,t about machine, le,arning is that the,re, is a lot of trial and e,rror involve,d in this study. be,fore, you se,le,ct the, right algorithm or structure,, you have, to try diffe,re,nt approache,s until you find the, right one,. in some, case,s, you might ne,e,d to use, more, than one, algorithm to ge,t the, right outcome,. as you try diffe,re,nt me,thods, e,nsure, your data is formatte,d and structure,d corre,ctly. With the, introductory knowle,dge, you gain from this book, you should be, able, to take, the, ne,xt ste,p in le,arning diffe,re,nt platforms and tools that will he,lp you with machine, le,arning mode,ling and training ai mode,s. some, of the, common visual tools you will use, include, microsoft azure, machine, le,arning studio, ibm watson studio, and google, cloud auto machine, le,arning. The, spe,cificity of the, proble,m you are, trying to solve, will also de,te,rmine, whe,the,r you succe,e,d in choosing the, right mode,l for machine, and de,e,p le,arning or not. you must cle,arly outline, the, proble,m you are, trying to solve, in orde,r to have, a be,tte,r chance, of mapping the, right mode,l for it. conside,r the, obje,ctive,s, nature, of data, and any othe,r factors that might affe,ct the, inte,nde,d re,sult whe,n choosing the, right algorithm for your work.  

CHAPTE ,R 1: MACHINE , LE ,ARNING: A BRIE ,F HISTORY   In the, mode,rn world of re,se,arch and busine,ss, machine, le,arning is one, subje,ct that come,s up all the, time,. this is a conce,pt that involve,s the, use, of ne,ural ne,twork mode,ls and algorithms to he,lp progre,ss computing syste,ms and boost the,ir pe,rformance, in diffe,re,nt auspice,s. algorithms play an important role, in machine, le,arning by he,lping de,ve,lope,rs cre,ate, arithme,tic mode,ls from basic data. the,se, mode,ls are, re,fe,rre,d to as training data. the, role, of training data is to he,lp the, machine, le,arning mode,ls inte,ract with data and make, de,cisions without the, de,ve,lope,r's programming mode,ls to make, the, de,cisions the,y do. Donald he,bb - the, organization of be,havior From the, e,xplanation above,, machine, le,arning is as close, an illustration of the, human brain as we, might come, across. machine, le,arning mode,ls are, ge,ne,rally de,signe,d to work in the, same, manne,r brain ce,lls inte,ract. the, history of machine, le,arning is litte,re,d with luminarie,s in the, fie,ld of compute,r and scie,ntific re,se,arch and de,ve,lopme,nt, and we, will start our historical ove,rvie,w in 1949, unde,r the, guidance, of donald he,bb. in his book,

the, organization of be,havior (he,bb, 1949), he, studie,d the, re,lationship in the, way ne,urons communicate, with one, anothe,r, and how the, conce,pt of e,xcite,d ne,urons make,s this possible,. He,bb studie,d the, re,lationship be,twe,e,n soma and axons in adjace,nt ce,lls in the, ne,urological ne,twork and notice,d that if one, ce,ll continuously he,lps the, ne,xt ce,ll ge,t fire,d up, its axon will de,ve,lop synaptic knobs that conne,ct with the, soma in the, adjace,nt ce,ll. the,se, obse,rvations forme,d the, foundation of studie,s in artificial ne,urons and artificial ne,ural ne,tworks. from his studie,s, scie,ntists furthe,r advance,d the,ir re,se,arch to sugge,st that it was possible, to influe,nce, the, re,lationship be,twe,e,n node,s (the,re,in obse,rve,d as ne,urons) and the, change,s that take, place, in e,ach ne,uron. furthe,r obse,rvation conside,ring two ne,urons re,ve,ale,d that whe,n activate,d at diffe,re,nt time,s, the,y had a we,ake,r re,lationship than whe,n we,re, activate,d simultane,ously. Samue,l arthur - ne,ural ne,tworks, che,cke,rs and rote, le,arning Thre,e, ye,ars afte,r he,bb’s studie,s, arthur samue,l, a re,se,arche,r at ibm, built a compute,r program that could play che,cke,rs. as you can imagine,, the, proce,ssing me,mory and re,source, capacity for compute,rs back in 1952 was limite,d. to mitigate, the, me,mory challe,nge,s, he, came, up with the, conce,pt of alpha-be,ta pruning (knuth e,t al., 1975). basically, he, de,signe,d a syste,m that would use, the, positions of individual pie,ce,s on the, che,cke,r's board to cre,ate, a scoring function. the, goal of this scoring function was to de,te,rmine, the, like,lihood of e,ithe,r of the, playe,rs winning the, game,, base,d on the,ir position. the, program samue,l built would use, a minimax strate,gy to de,te,rmine, the, be,st possible, move, (sackrowitze,t al., 1986). this program

would furthe,r advance,d into what we, curre,ntly ide,ntify as the, minimax algorithm. Samue,l re,alize,d the, ne,e,d to advance, his program to adapt to diffe,re,nt playing e,ncounte,rs, he,nce, he, introduce,d more, te,chnique,s to improve, it, an approach that he, re,fe,rre,d to as rote, le,arning (hoosain, 1970). in this conce,pt, the, program would re,cord and re,call e,ve,ry position it he,ld pre,viously, the, positions it had se,e,n and factor in the, value, of the, re,wards. it was around this time, in 1952 that he, coine,d the, phrase, machine, le,arning. Rose,nblatt’s pe,rce,ptron Othe,r e,xpe,rts we,re, ke,e,n to advance, the, ide,as propose,d by samue,l and he,bb. in 1957, frank rose,nblatt built on the,ir studie,s on the, e,fforts of machine, le,arning and the, brain ce,ll inte,raction re,spe,ctive,ly to cre,ate, what he, re,fe,rre,d to as the, pe,rce,ptron (rose,nblatt, 1958). the, inte,re,sting thing about the, pe,rce,ptron is that while, most pe,ople, came, to inte,ract with it as a program, rose,nblatt me,ant for it to be, a machine,. He, built the, program as an image, re,cognition program for the, ibm 704. in as far as scalability is conce,rne,d, rose,nblatt cre,ate,d algorithms for the, pe,rce,ptron that could be, use,d with othe,r machine,s. the, pe,rce,ptron would be, re,cognize,d as the, first ne,uro-compute,r that was succe,ssfully de,ploye,d. While, the, ide,a was a good one,, rose,nblatt e,xpe,rie,nce,d a lot of challe,nge,s in de,ployme,nt. the, pe,rce,ptron was a promising proje,ct, but it ne,ve,r succe,e,de,d in ide,ntifying face,s or most visual patte,rns that could he,lp in the, distinction and ide,ntification of individuals. as a re,sult of this disappointme,nt and the, inability to source, additional funding to advance, the, proje,ct furthe,r, the, ne,ural ne,twork re,se,arch stalle,d. re,se,arch on machine, le,arning and ne,ural ne,tworks ge,ne,rally quie,te,d down until the, 1990s.

Marce,llo pe,lillo - the, ne,are,st ne,ighbor algorithm Fast forward to 1967, patte,rn re,cognition as use,d in machine, le,arning today came, to light unde,r the, ne,are,st ne,ighbor algorithm. the, ne,are,st ne,ighbor algorithm was one, of the, first algorithms that was imple,me,nte,d in a bid to he,lp sale,spe,ople, find the, be,st possible, route,s. sale,spe,ople, ge,ne,rally trave,le,d a lot, and a suitable, route, that me,ant spe,nding le,ss time, trave,ling was an ide,al re,comme,ndation. The, algorithm was introduce,d to make, trave,l more, e,fficie,nt for sale,spe,ople,. through this algorithm, the, use,r would choose, the,ir pre,fe,rre,d city the,n have, the, algorithm che,ck all the, citie,s close,st to the, one, the,y chose, until all the,y visite,d all the, citie,s. Pe,rce,ptrons and multilaye,rs The,re, was a ne,e,d for more, proce,ssing powe,r give,n how fast machine, le,arning was advancing and the, prospe,cts for the, future,. re,se,arch in ne,ural ne,tworks was just picking up in the, 1960s. re,se,arche,rs re,alize,d that whe,n the,y use,d one, pe,rce,ptron, the, machine,s had lowe,r proce,ssing capacity than whe,n the,y use,d more, than one, pe,rce,ptron. this was afte,r trials and te,sts with multilaye,rs. from the,se, findings, more, studie,s into ne,ural ne,tworks we,re, conducte,d. The, use, of multilaye,rs late,r gave, birth to backpropagation and fe,e,d-forward ne,ural ne,tworks. in backpropagation, re,se,arche,rs built ne,tworks that could automatically adjust the,ir node,s and ne,urons, in the, proce,ss of adapting to diffe,re,nt e,xpe,rie,nce,s (bod, 2001). one, of the, be,st e,xample,s of this was backward e,rror propagation. in this case,, output e,rrors could be, trace,d back

to the, ne,twork laye,rs to unde,rstand the,ir nature, and origin. at the, mome,nt, backpropagation plays an important role, in training de,e,p ne,ural ne,tworks. While, the,re, was a lot of promise, to the, use, of pe,rce,ptrons, the,y prove,d futile, in handling complicate,d assignme,nts. be,cause, of this re,ason, artificial ne,ural ne,tworks we,re, introduce,d. the,y had ste,alth laye,rs that spe,cifically handle,d this issue,. artificial ne,ural ne,tworks have, since, be,come, one, of the, important tools in machine, le,arning. basically, to use, a ne,ural ne,twork you ne,e,d input and output parame,te,rs. the,se, are, se,rve,d by the, input and output laye,rs, alongside, hidde,n laye,rs that he,lp in data conve,rsion be,twe,e,n the, input and output proce,sse,s. the, role, of the, hidde,n laye,rs is to proce,ss data that is too complicate,d for e,ve,n the, be,st human programme,r to handle,. the,y he,lp in ide,ntifying complicate,d patte,rns and tre,nds. unlike, othe,r proce,sse,s and le,arning me,thods, it is impossible, for any human to te,ach the,se, laye,rs ne,w patte,rns be,cause, we, cannot handle, the, comple,xity be,hind the,m. Going se,parate, ways For the, longe,st time,, machine, le,arning and artificial inte,llige,nce, have, always be,e,n discusse,d in the, same, light. howe,ve,r, this is not suppose,d to be, the, case,. while, the, discipline,s share, some, commonalitie,s, the,ir dime,nsional focus is diffe,re,nt. this was e,vide,nt during the, 70s and 80s. up until that time,, machine, le,arning was one, of the, training module,s use,d for e,mpowe,ring artificial inte,llige,nce,. on its part, artificial inte,llige,nce, was advancing away from the, use, of algorithms to dwe,ll on le,arning through proce,sse,s that involve,d knowle,dge, and logical ope,rations. E,xpe,rts in compute,r scie,nce, and re,se,arch in artificial inte,llige,nce, e,ve,ntually quit working on ne,ural ne,twork re,se,arch. this rift be,twe,e,n artificial inte,llige,nce, and machine, le,arning le,d most machine, le,arning e,xpe,rts and re,se,arche,rs to re,focus the, dynamics of the,ir work to providing

solutions to re,al-life, proble,ms inste,ad of advancing artificial inte,llige,nce, obje,ctive,s. Inste,ad of using the, me,thods advance,d in artificial inte,llige,nce, studie,s, machine, le,arning e,xpe,rts inve,ste,d he,avily in how to use, statistical and probabilistic approache,s in proble,m-solving. ne,ural ne,tworks we,re, once, again an important part of the, re,se,arch proce,ss and this he,lpe,d the, studie,s in machine, le,arning thrive, through the, 90s. we, must also re,cognize, the, fact that while, the,se, studie,s we,re, going on, the, growth of the, inte,rne,t was also taking shape,. data upon which the, mode,ls could train was incre,asingly available,, and the, e,ase, of sharing information online, also he,lpe,d this cause,. Robe,rt schapire, - the, stre,ngth of we,ak le,arnability One, of the, most important mile,stone,s in the, history of machine, le,arning was boosting. boosting is a proce,dure, whe,re, spe,cific algorithms are, use,d to e,liminate, possible, bias in supe,rvise,d le,arning approache,s. boosting algorithms ge,ne,rally he,lp to improve, and stre,ngthe,n we,ak le,arne,rs. this ide,a was introduce,d in the, stre,ngth of we,ak le,arnability (schapire,, 1990). in his work, he, obse,rve,d that it was possible, to build a strong le,arne,r from a numbe,r of we,ak le,arne,rs. we,ak le,arne,rs, in this case,, re,fe,rre,d to classifie,rs that share, a slight corre,lation with the, true, classification, unlike, strong le,arne,rs that are, prope,rly aligne,d. The, boosting algorithms are, basically a composition of se,ve,ral we,ak classifie,rs that compile, to form a strong classifie,r. once, the,y are, compounde,d into a strong classifie,r, the, accuracy of the, le,arne,rs is de,te,rmine,d by we,ighting. the,re, are, many type,s of boosting algorithms. what se,ts the,m apart is the, me,thod use,d in training the, we,ighte,d data

points. one, of the, most popular machine, le,arning algorithms today, adaboost, is one, such e,xample,. adaboost has constantly prove,n ade,quate, in working with we,ak le,arne,rs. othe,r boosting algorithms: ● totalboost ● madaboost ● brownboost ● xgboost ● lpboost ● logitboost All the,se, algorithms are, supporte,d and work inside, the, anyboost e,nvironme,nt. Advancing into spe,e,ch and facial re,cognition Most of the, advance,me,nt we, have, e,xpe,rie,nce,d in spe,e,ch re,cognition at the, mome,nt is thanks to long short-te,rm me,mory (lstm) (hochre,ite,r, e,t al., 1997). this is a ne,ural ne,twork te,chnique, that capture,s e,ve,nts that happe,ne,d many discre,te, ste,ps be,fore, the, curre,nt e,ve,nt. lstm can re,tain me,mory for thousands of e,ve,nts pre,ce,ding the, curre,nt e,ve,nt, a te,chnique, that is ne,ce,ssary for de,ve,loping and advancing spe,e,ch re,cognition programs. The,re, have, be,e,n othe,r spe,e,ch re,cognition programs in the, marke,t, but none, we,re, as prolific as lstm. by the, ye,ar 2007, lstm was mile,s ahe,ad of most spe,e,ch re,cognition tools, programs and software, available,. google, took advantage, of this in 2015 and improve,d the,ir spe,e,ch re,cognition algorithms by imple,me,nting an lstm that was traine,d through conne,ctionist

te,mporal classification (ctc). as a re,sult, the, google, spe,e,ch re,cognition algorithm improve,d in e,fficie,ncy and pe,rformance, by more, than 45%. Succe,ss in spe,e,ch re,cognition would soon be, transfe,rre,d to facial re,cognition. the, national institute, of standards and te,chnology program he,ld a facial re,cognition grand challe,nge, in 2006 to te,st the, most popular algorithms in this dime,nsion. among the, fe,ature,s te,ste,d we,re, high-re,s facial photos, iris scans and image,s, and 3d facial scans. by the, e,nd of this e,ve,nt, it was e,vide,nt that algorithms that had be,e,n in use, since, the, e,arly 2000s we,re, no match for the, mode,rn facial re,cognition algorithms. most of the,m we,re, outpe,rforme,d more, than 10-100 time,s. some, unique, algorithms could pe,rform be,tte,r than individual use,rs, and could e,ve,n te,ll apart ide,ntical twins. By 2012, many of the, le,ading te,ch companie,s we,re, alre,ady e,xpe,rime,nting with machine, le,arning in diffe,re,nt re,spe,cts. e,xpe,rts at google,’s x lab built an algorithm in 2012 that would browse, the, inte,rne,t on its own and find cat vide,os. face,book followe,d suit two ye,ars late,r with de,e,pface, (taigman e,t al., 2014). this algorithm could accurate,ly ide,ntify pe,ople, in photos as accurate,ly as individual use,rs would. Pre,se,nt day machine, le,arning Ge,tting compute,rs to pe,rform tasks without supe,rvision is one, of the, highlights of machine, le,arning. while, the, mode,ls can be, programme,d to carry out an assignme,nt at de,ve,lopme,nt le,ve,l, whe,n in ope,ration the,y are, not e,xplicitly instructe,d in any way. the, machine,s le,arn from inte,raction with diffe,re,nt use,rs and use,r patte,rns/be,havior, and infe,re,ncing input and output data fe,d into the, syste,m. Curre,ntly, we, are, living through some, amazing te,chnological advance,me,nts, all made, possible, through machine, le,arning. a lot of

te,chnique,s and te,chnologie,s have, since, be,e,n advance,d through machine, le,arning, ushe,ring in a ne,w dime,nsion in machine, le,arning and artificial inte,llige,nce, as we, he,ad into the, unce,rtain future,. This discipline, has grown and will ke,e,p growing in e,pic stride,s e,spe,cially whe,n we, look at the, prospe,ct for analytics, the, inte,rne,t of things and robotics, to me,ntion some, te,ch that will shape, our future,. be,low are, some, of the, common instance,s whe,re, we, inte,ract with machine, le,arning mode,ls from time, to time,: ● conve,rsations with othe,r pe,ople, online, through natural language, proce,ssing (nlp) ● de,mand or ne,e,d-base,d fle,xibility in pricing through dynamic pricing algorithms ● de,cision-making programs that use, le,arning manage,me,nt syste,ms ● pe,rsonalization of custome,r product re,comme,ndations ● ide,ntifying patte,rn change,s in use,r activitie,s, he,nce, e,mpowe,ring fraud de,te,ction age,ncie,s ● stre,amline,d and re,al-time, data analysis By de,sign, machine, le,arning mode,ls are, built to le,arn infinite,ly. this is possible, by inte,racting with diffe,re,nt kinds of data and updating the,ir syste,ms accordingly. be,cause, of this, the, mode,ls be,come, more, e,fficie,nt and accurate, e,ach time, the,y acce,ss ne,w data or inte,ract with ne,w use,rs. the,se, mode,ls are, built for e,xte,nsibility, such that the,y don’t buckle, unde,r pre,ssure, from ne,w data, he,nce, supporting the, e,fficie,ncy obje,ctive,. machine, le,arning algorithms and mode,ls are, curre,ntly use,d to manage, many comple,x ope,rations in busine,ss e,nvironme,nts. with this te,chnology, we, can pre,dict anything from a possible, dise,ase, outbre,ak to price, fluctuations at the, stock e,xchange,.